Researching creator-platform buyers and creators without mixing the two populations

Useful creator discovery platforms audience research begins with a recent decision, not a social profile. Recruit people because they can evidence a procurement, discovery, campaign, data-governance or creator-participation event. Their job title, follower count or assumed interest does not establish eligibility.

This England-focused guide designs research without inventing participants or findings. It keeps qualitative accounts within the recruited sample and refuses to turn them into estimates of national demand, adoption or creator behaviour.

What to take away

Identify each population before recruitment

"Audience" can hide several groups whose experiences cannot be combined. Define them in the protocol:

Acceptable evidence for eligibility could be a redacted decision log, invitation, account screen or contract section.

Typical planning figures are 12 to 20 qualitative interviews per population, with a minimum of 8 where the decision path is narrow. A survey might use quotas of about 100 buyers and 200 creators, subject to frame size and precision needs.

The same person may hold several roles, but the analysis should not merge them. A supplier employee cannot speak for an indexed creator. An advertiser user may not control procurement. A talent manager's statement does not establish a creator's consent or authority unless the underlying record does.

Set a recent-event eligibility rule

Write a pre-outreach screening window. A buyer participant may need documentary involvement in a platform selection, contract renewal, manual creator search or governed campaign hand-off completed in the last 6 to 12 months.

A creator participant may need direct experience of joining a marketplace, discovering that a record was indexed, requesting a correction or receiving a permissioned campaign approach through a specified route.

Ask what record can verify the event without requiring unnecessary confidential data. Never screen on inferred age, residence, ethnicity, health, religion, sexuality, interests, audience composition, authenticity, suitability or buying intention.

The GOV.UK Service Manual advises researchers to define participant groups and recruitment criteria, check for access needs and consider recruitment bias. It is government service-research guidance rather than a rule for private commercial studies, so a named research-methods reviewer must adapt it to this project.

Build separate frames for buyers and creators

For buyer organisations, construct a declared frame using eligible English operating units, a relevant procurement population or a permissioned customer register. The ONS 2025 business dataset can support broad regional, size and industry strata among VAT or PAYE registered businesses. It does not reveal creator-platform use, purchasing authority or campaign activity.

Named routes include Companies House, the ONS UK business dataset and the Market Research Society's Company Partner directory; the first two cover the whole UK, so England-only strata must be drawn explicitly.

Creator research needs a different frame.

For creators, Arts Council England's funded networks cover England only and can provide a permissioned route, though platform opt-in lists remain the primary source.

Use direct applications, a permissioned membership list, a documented platform notice or another reviewed route. Do not scrape profiles to recruit people. Do not treat a public email address or direct-message function as research permission.

Record the frame date, source, coverage, duplicates, unavailable records and exclusions. If creators without public profiles, people using agents, former users or those who exercised deletion rights cannot enter the frame, describe that coverage defect beside the findings.

Keep participation, processing and later contact apart

The GOV.UK guidance on informed consent for user research covers the information participants should receive about recording, sharing, retention, withdrawal and data control. Consent to take part is one decision.

The ICO's research-processing guidance explains that ethical participation consent is distinct from consent used as a UK GDPR lawful basis. A named privacy reviewer must document the actual purpose, roles, lawful basis, special-category conditions where applicable, minimisation, transparency, rights, retention and transfer position.

Later sales contact is another decision. The ICO says genuine market research is outside direct marketing, but a survey containing promotional material or collecting details for later campaigns may fall inside it, in its PECR overview. Keep the research invitation, participation record and any future marketing permission in separate fields and systems.

Design the fieldwork record before asking questions

Give every invitation and screener a version number. Capture its issue date, channel, sender, source list, eligibility rule and suppression process. For each role and stratum, report:

Incentives or other consideration require an amount, currency, conditions, payment owner and conflict note. A reward should not be hidden inside campaign recruitment, a testimonial request or future platform access. A safeguarding lead must approve any route where under-18 status may arise; researchers should not infer age from content or appearance.

Ask participants what accessible format or adjustment they need. Government guidance on involving disabled people in social research covers inclusive design and accessible fieldwork, though it was written for government research. Test the exact invitation, consent material, interview mode and survey journey with relevant users.

Interview the recent task, not a hypothetical preference

Start with a dated event: "Show the last creator-discovery decision you were authorised to make." Ask what initiated it, which records were available, who approved each hand-off, what was missing and what stopped or changed the process. Do not ask whether a participant "likes influencer tools" or would buy an imagined product.

Buyer interviews can reveal approval, evidence and exit requirements. Participating creators can describe information they provided and controls they saw. Indexed creators can report what they observed in their own record. None should be invited to guess how other creators behave or how a platform built a field.

Remove statements about third parties unless they are necessary and permissioned. A participant's account of a supplier is testimony about that experience, not verified product capability. Seek a dated first-party service record and independent evidence before publishing a technical claim.

Treat personas as compact evidence records

A creator-platform persona should represent a recurring decision pattern within a declared sample. It is not a fictional person. Do not add a name, photograph, invented quotation, demographic backstory or composite creator profile.

Each record needs a population, eligible cases, evidence period, decision, constraints, observed behaviours, counter-evidence, missing perspectives, owner and expiry. A worked record might look like this:

It should not become "the marketing manager persona" unless every underlying case and variation supports that label.

Set a retirement rule. Withdraw the persona if new interviews contradict its decision path, if the service edition changes, if the sample no longer covers the target population, or if the evidence is too sparse to protect confidentiality.

Compare survey routes, not brand promises

Manual route: approved invitation register, accessible questionnaire in an existing buyer-controlled system, documented export, compare the organisation's already approved survey service with procuring a new product.

Name products only if current exact-edition records cover service scope, privacy, roles, storage and transfers. They must also cover security, accessibility, export, deletion and support. Cover contract terms and UK pricing where price is used.

The Government Analysis Function's questionnaire design guidance recommends attention to ethics, respondent burden, accessible design and qualitative testing. It also distinguishes a questionnaire from the wider survey method. A polished form cannot repair a weak frame, leading questions or hidden non-response.

Test with synthetic data before collecting personal information. Verify routing, validation, accessible use, duplicate handling, partial saves, withdrawal, export, suppression and deletion. Unknown supplier fields remain unknown and become procurement questions, not favourable assumptions.

Research competitors without inventing a market

Create separate evidence layers. Companies House records can support legal identity for incorporated UK suppliers. A supplier service page supports an attributed description on its stated date. A directory entry proves only that the listing existed. Public case studies and testimonials remain publisher or supplier claims unless the underlying parties and evidence are verified.

The CMA's current reviews and endorsements collection addresses genuine and accurate review information and transparent endorsements. It does not validate a competitor's clients or service quality.

Do not infer platform coverage, creator counts, record accuracy, popularity, revenue, market share, reach or campaign results from search position, roster pages, logos, testimonials or social activity. Also separate a discovery database from a marketplace, workflow platform, affiliate service, talent manager and managed agency.

Analyse within the limits of the design

Qualitative themes describe the recruited participants and their contexts. Count cases supporting and contradicting each theme, but do not report percentages as England prevalence. A buyer survey needs a probability frame or defensible inference design, transparent weighting and uncertainty before it can speak beyond respondents.

The UK Statistics Authority's Code of Practice for Statistics says data sources, limitations, potential bias and uncertainty should be explained. This commercial study would not automatically be an official statistic, yet those quality principles are a sound publication discipline.

Report non-response, item missingness, proxy answers, recall period, mode effects and conflicts. Segment only by pre-declared operational variables that were collected lawfully and have enough cases to protect people.

Turn evidence into a reversible decision

Before fieldwork, write the decision that each finding could change. A platform may remain HOLD if buyers cannot show a recent task, indexed creators cannot access correction routes, a supplier cannot explain field provenance, or the research sample excludes the people most affected.

Store a decision log containing the evidence version, interpretation, contrary cases, owner, threshold, action and review date. Retest after supplier, platform, privacy-notice, export or campaign-process changes. Delete recruitment data on the approved schedule and propagate valid corrections or withdrawals to research copies.

This article remains on publication hold. It needs named UK research-methods, statistical, privacy and PECR, safeguarding, accessibility, IP, security, advertising, consumer and platform-procurement review, plus publication-day URL and guidance checks.

Before you act

Common questions

What evidence makes someone eligible for this research?

The first section sets out the evidence needed for each population.

Why can buyer and creator findings not be combined?

Audience can hide several groups whose experiences cannot be combined, including buyer organisations, advertiser users, procurement authorities, platform operators, intermediaries and creators. A supplier employee cannot speak for an indexed creator.

How should personas be used in this research?

A creator-platform persona should represent a recurring decision pattern within a declared sample, not a fictional person. Do not add a name, photograph, invented quotation, demographic backstory or composite profile. Each record needs a population, eligible cases, evidence period, decision, constraints, observed behaviours, counter-evidence, missing perspectives, owner and expiry.